YouGov suppressed polling in last few weeks of the 2017 general election campaign because it was too positive about Labour, a senior former employee has said.
Chris Curtis, a former political research manager, said “everybody panicked at the backlash” to a YouGov poll suggesting there could be a hung parliament, and the firm subsequently drew back from publishing some data that showed Jeremy Corbyn’s Labour doing well. YouGov denies the allegation.
At the time, the Conservative MP Nadhim Zahawi, a cofounder of YouGov and now the education secretary, phoned the company’s chief executive, Stephan Shakespeare, and said he would call for his resignation if the latter poll was wrong, according to an account in a book by the political journalist Tim Shipman.
Curtis said the furore meant “our polling and coverage was a lot worse for the rest
of the campaign”, with some polling on a leadership debate showing a favourable response to Corbyn not published.
“We did a fantastic debate poll in the hours following the debate that Corbyn took part in. The results were stark – Corbyn won by a country mile, and one in four Tory voters thought he was best,” he wrote in a Twitter thread.
“But despite having written the story and designed the charts, we were banned from releasing the story because it was too positive about Labour.
“Similarly, there were a few ‘minor’ methodology changes for the final poll which increase the Tory lead. This was done after pressure from high-ups (and despite protests from those of us who thought it wasn’t OK).”
Curtis, who now works for another polling firm, Opinium, said he had learned a few lessons: “1) Trust the data 2) Ignore political commentators when they don’t trust the data 3) The electorate is volatile now, bigger shifts are to be expected.”
YouGov said the allegation was “incorrect”. A spokesperson said: “There was a poll run by Chris following the debate in Cambridge on 31 May 2017. When reviewed by others in the YouGov political team, it was clear that the sample of people who watched the debate significantly over-represented Labour voters from the previous election.
“We take our responsibilities as a research organisation seriously and we could not have published a poll from a skewed sample that favoured any party. No serious polling organisation would have published this.
“The idea that YouGov would suppress a poll that was ‘too positive about Labour’ is plainly wrong – as evidenced by the fact that in the 2017 election YouGov published an MRP model showing Labour doing significantly better compared to most other polling organisations.”
However, Curtis stood by his comments. He said he believed the data was not published because the company was nervous about the consequences of being wrong, rather than because Zahawi had any hand in the situation.
“Given people are asking, I’m not saying that Nadhim directly had a hand in YouGov changing its results to suit Tory ends (as some have interpreted) but pointing to the general sense of panic YouGov had at the time which led to certain decisions,” he said.
Curtis said the methodology of the poll was the same as all other polling that YouGov had done, and that the overall sample would have been weighted to be representative of the population.
He said it was possible that more Labour voters had watched the debate than Conservative voters.
But he added: “Either way, the most important finding of the poll, the one I wanted to focus on and thought was most important, was that a good chunk of Tory voters thought Corbyn had won. This is rare in a debate poll where results normally fall down party lines.”
Zahawi said: “This was clearly a joke between two good friends, who had previously been business partners for several years.
“Stephan continues to be one of my closest friends and at no point since leaving YouGov in 2010 have I had any influence on the company. Suggesting otherwise is untrue.”
• This article was amended on 11 June 2022 to add an update to the sub-heading. As the Guardian reported on 10 June 2022, Chris Curtis retracted his comments the day after this article was published.
